特征提取及数据扩充的GA-LightGBM半导体质量检测方法

GA-LightGBM method of semiconductor quality inspection based on feature extraction and data expansion

半导体质量检测数据具有的“相关性、冗余性、不平衡性”等特点,导致传统的分类算法效率较低,为此,提出一种基于特征提取及数据扩充的GA-LightGBM(genetic algorithm-light gradient boosting machine)质量检测方法。通过结合主成分分析(principal component analysis,PCA)、合成少数类过采样技术(synthetic minority oversampling technique,SMOTE)、遗传算法和LightGBM这4种方法,实现对产品质量的有效识别。实验结果表明,相较于传统分类算法,提出的方法可以有效提升质量检测的效率。

Semiconductor quality inspection data exhibit characteristics such as correlation,redundancy,and imbalance,which lead to lower efficiency in traditional classification algorithms.To address this challenge,we propose a quality inspection method named GA-LightGBM(genetic algorithm-light gradient boosting machine)that leverages feature extraction and data augmentation techniques.This approach combines principal component analysis(PCA),synthetic minority oversampling technique(SMOTE),genetic algorithm,and LightGBM.Experimental results demonstrate that,compared to traditional classification algorithms,the proposed method significantly improves the efficiency of quality inspection.